1. Field of the Invention
The present invention relates to communication systems, and, in particular, to adaptive equalization using pattern recognition.
2. Description of the Related Art
Fibre Channel (FC) is a high performance communications interconnect standard specifying data transfer that is capable of transporting a relatively great amount of user data traffic between, for example, desktop workstations, mass storage devices, peripheral components, and host systems at rates much faster that those present in typical local area network (LAN) communication networks. FC is rapidly becoming a generally accepted network architecture for stage area networks (SANs), which are increasingly important for managing the large volume and complexity of data in post internet-era applications.
One component of FC devices is a serializer and de-serializer (SerDes) device that is commonly used in high speed communications to convert data between serial and parallel interfaces in each transmit/receive direction. A common coding scheme used with SerDes devices, such as those employed in FC, is (8B10B) coding. All information is first encoded 8-bits at a time into a 10-bit transmit character. The 10-bit character is then sent serially bit-by-bit over a communication link, such as an electrical or optical link. This (8B10B) encoding supports DC-balance, provides framing, and guarantees signal transitions. Guaranteed transitions allow a receiver to extract the embedded clock signal (clock data recovery, or CDR), while control codes allow framing, typically on the start of a data packet. This encoding scheme also improves error detection with running disparity, providing separation of data bits from control bits, and permits derivation of byte and word synchronization. The typical (8B10B) SerDes parallel side data bus interface has 1 clock line and 8 (or multiples of 8) data lines for each transmit and receive lane.
SerDes systems transfer data, but during some periods data might not be sent over the communication link: During these non-data periods, many standards, including FC, specify that a primitive signal be employed to keep a low-energy signal with transitions in the link, allowing for continued timing extraction and reduction of sample-timing wander at the front end of the receiver. By keeping a receiver's front end active, the receiver can transition to active data detection relatively rapidly while maintaining a low bit error rate (BER).
An important signal degradation that occurs in SerDes devices is from electromagnetic interference (EMI) introduced by toggling signals, which EMI degradation increases as data rates increase. Generation of EMI is closely regulated by the FCC (Federal Communications Commission) since it can interfere with other user devices. Consequently, lowering the EMI of a device might lower device costs of a given implementation by lowing shielding requirements.
To reduce lower EMI while minimizing impact to existing designs, the T11 Technical Committee has proposed replacing the currently employed primitive signal IDLE with an ARBFF signal for synchronization. The ARBFF signal exhibits a lower transition density than the IDLE signal. Both an IDLE signal and an ARBFF signal are specified as an “ordered set” of signals where each element of the ordered set exhibits a distinct pattern. Table 1 summarizes the FC data rates and proposed primitive signals for those data rates.
TABLE 1BeginningRunningFC StandardSerial Data RatePrimitive SignalDisparityOrdered Set1G 1.0625 GbpsIDLENegativeK28.5-D21.4-D21.5-D21.52G  2.125 GbpsIDLENegativeK28.5-D21.4-D21.5-D21.54G  4.25 GbpsIDLENegativeK28.5-D21.4-D21.5-D21.58G   8.5 GbpsARBFFNegativeK28.5-D20.4-D31.7-D31.710G 10.51875 GbpsARBFFNegativeK28.5-D20.4-D31.7-D31.7
Lower transitional density of the ARBFF signal's repeating data pattern, however, works against adaptive equalization (AEQ) by the receiver, and SerDes receiver performance in general. The effect is amplified when data traffic resumes after a long period of link synchronization. During the long period of non-random repeating data patterns and with low transitional density, AEQ might drifts away from relatively optimum settings for random data detection, leading to incorrect data decisions or data corruption once user data traffic resumes.